IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Helix vs. Lame, Which one to choose at higher bitrates?
Gainless
post Dec 31 2012, 04:21
Post #1





Group: Members
Posts: 169
Joined: 28-October 11
Member No.: 94764



So I've seen that Helix was rated equal if not superior to Lame at the last mp3 listening test. I've done a few tests with Helix on my own at 192 kbps with the result that it does a lot better than Lame on many "killer" samples. However there are issues with the handling of high frequencies at bass heavy material, e.g. that quite audible content is removed when a kick appears (was also noticed in this thread), which seems to be little dependent on how high the choosen bitrate/quality is.
From that perspective alone I would prefer Helix, also regarding the encoding speed, but I don't know if this is the only "Achilles heel" of this encoder. Are there any other reasons than the performance at high frequencies to choose Lame qualitywise over Helix?

This post has been edited by Gainless: Dec 31 2012, 04:41
Go to the top of the page
+Quote Post
BFG
post Dec 31 2012, 05:16
Post #2





Group: Members
Posts: 206
Joined: 22-July 12
Member No.: 101637



QUOTE (Gainless @ Dec 30 2012, 21:21) *
So I've seen that Helix was rated equal if not superior to Lame at the last mp3 listening test. I've done a few tests with Helix on my own at 192 kbps with the result that it does a lot better than Lame on many "killer" samples. However there are issues with the handling of high frequencies at bass heavy material, e.g. that quite audible content is removed when a kick appears...

Which mp3 listening test are you referring to? I'd be interested in seeing the results - particularly since my hearing isn't good enough to ABX even so-called "killer samples" at high bitrates on either player.

That said. I've been using halb27's 3.100h LAME variant and have been very happy with the results. LAME 3.100 is an alpha build that is reported to resolve most of the preecho issues with earlier versions of LAME. And halb27's h variant forces higher bitrates in certain situations, which typically resolves temporal or tonal issues.
Go to the top of the page
+Quote Post
Gainless
post Dec 31 2012, 11:31
Post #3





Group: Members
Posts: 169
Joined: 28-October 11
Member No.: 94764



You can find the listening test here. The killer samples I've used are mostly easy ones to catch, as Lame is introducing obvious distortions/glitches at the midrange there, haven't checked them with Halb's extension yet, though.

This post has been edited by Gainless: Dec 31 2012, 11:32
Go to the top of the page
+Quote Post
halb27
post Dec 31 2012, 13:39
Post #4





Group: Members
Posts: 2425
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



QUOTE (Gainless @ Dec 31 2012, 11:31) *
You can find the listening test here. The killer samples I've used are mostly easy ones to catch, as Lame is introducing obvious distortions/glitches at the midrange there, haven't checked them with Halb's extension yet, though.

If you want to try, it would be very interesting to learn about Lame 3.100a2's behavior which has the potential to improve problems. Sure the results of my extension are interesting as well.


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
Gainless
post Dec 31 2012, 14:30
Post #5





Group: Members
Posts: 169
Joined: 28-October 11
Member No.: 94764



QUOTE (halb27 @ Dec 31 2012, 13:39) *
If you want to try, it would be very interesting to learn about Lame 3.100a2's behavior which has the potential to improve problems. Sure the results of my extension are interesting as well.

They are all present here on HA, not that many ones from myself though:
Sycho Active
Meduzz (Pre-echos)
Poquito Mas (Already included in the sample section of 3.100h)

I compared V2 of 3.100a2 and V2+ of your 3.100h extension with these, with the result that it performs better on the last two but not really much on the first one, despite the bitrate boost.

This post has been edited by Gainless: Dec 31 2012, 14:31
Go to the top of the page
+Quote Post
halb27
post Dec 31 2012, 18:15
Post #6





Group: Members
Posts: 2425
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



Thanks for testing.
Does Helix provide a better quality for these samples?


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
Wombat
post Dec 31 2012, 19:39
Post #7





Group: Members
Posts: 1011
Joined: 7-October 01
Member No.: 235



QUOTE (Gainless @ Dec 31 2012, 05:21) *
From that perspective alone I would prefer Helix, also regarding the encoding speed, but I don't know if this is the only "Achilles heel" of this encoder. Are there any other reasons than the performance at high frequencies to choose Lame qualitywise over Helix?

What i like on lame is that gapless works really well. Helix has no delay/padding info for playback.
Go to the top of the page
+Quote Post
Gainless
post Dec 31 2012, 22:33
Post #8





Group: Members
Posts: 169
Joined: 28-October 11
Member No.: 94764



QUOTE (halb27 @ Dec 31 2012, 18:15) *
Thanks for testing.
Does Helix provide a better quality for these samples?

Helix did worse at the Sycho Active sample, but was on par/better at the rest with -V120/192kbps, at least to my ears.

QUOTE (Wombat @ Dec 31 2012, 19:39) *
QUOTE (Gainless @ Dec 31 2012, 05:21) *
From that perspective alone I would prefer Helix, also regarding the encoding speed, but I don't know if this is the only "Achilles heel" of this encoder. Are there any other reasons than the performance at high frequencies to choose Lame qualitywise over Helix?

What i like on lame is that gapless works really well. Helix has no delay/padding info for playback.

That's fortunately not an issue for me as a non-audiobook-listener, but still a good point.

This post has been edited by Gainless: Dec 31 2012, 22:33
Go to the top of the page
+Quote Post
DonP
post Dec 31 2012, 23:42
Post #9





Group: Members (Donating)
Posts: 1471
Joined: 11-February 03
From: Vermont
Member No.: 4955



QUOTE (Gainless @ Dec 31 2012, 16:33) *
QUOTE (halb27 @ Dec 31 2012, 18:15) *

What i like on lame is that gapless works really well. Helix has no delay/padding info for playback.

That's fortunately not an issue for me as a non-audiobook-listener, but still a good point.


It's an issue for me as a music listener.
Go to the top of the page
+Quote Post
Gainless
post Jan 1 2013, 03:49
Post #10





Group: Members
Posts: 169
Joined: 28-October 11
Member No.: 94764



You mean like fades between songs on an album? Forgot about that, why didn't get it fixed btw?
Go to the top of the page
+Quote Post
BFG
post Jan 1 2013, 07:44
Post #11





Group: Members
Posts: 206
Joined: 22-July 12
Member No.: 101637



This thread makes me wonder if the LAME and Helix teams have considered joining forces. Why have two competing encoders when you can have one that is the best of both worlds?
Go to the top of the page
+Quote Post
saratoga
post Jan 1 2013, 08:54
Post #12





Group: Members
Posts: 4905
Joined: 2-September 02
Member No.: 3264



QUOTE (BFG @ Jan 1 2013, 01:44) *
This thread makes me wonder if the LAME and Helix teams have considered joining forces. Why have two competing encoders when you can have one that is the best of both worlds?


Generally speaking, the challenges of merging two very different approaches to the same problem are often greater then just improving each individually.
Go to the top of the page
+Quote Post
db1989
post Jan 1 2013, 12:22
Post #13





Group: Super Moderator
Posts: 5275
Joined: 23-June 06
Member No.: 32180



QUOTE (Gainless @ Jan 1 2013, 02:49) *
You mean like fades between songs on an album?
Not fades, just gaps. As the name implies. wink.gif
Go to the top of the page
+Quote Post
BFG
post Jan 2 2013, 01:19
Post #14





Group: Members
Posts: 206
Joined: 22-July 12
Member No.: 101637



QUOTE (saratoga @ Jan 1 2013, 01:54) *
Generally speaking, the challenges of merging two very different approaches to the same problem are often greater then just improving each individually.

That's probably true. Still, if Helix and LAME both use relatively similar principles - i.e. if they both use a psychoacoustic model - it might be possible to conjure up a program which encodes each frame under both models and then keeps the "most accurate" frame.

And yes, that may make no sense whatsoever given the technical details of how the two work.
Go to the top of the page
+Quote Post
IgorC
post Jan 2 2013, 03:15
Post #15





Group: Members
Posts: 1553
Joined: 3-January 05
From: ARG/RUS
Member No.: 18803



While one MP3 encoder puts a short block another will put long block. Both can have good reasons to do so.
You can't get the best of both because it will imply to go with short and long blocks at the same time.

Some recent codecs can use long and short blocks at the same time at different frequency bands. But MP3 format doesn't allow it.
Go to the top of the page
+Quote Post
Gainless
post Jan 5 2013, 16:54
Post #16





Group: Members
Posts: 169
Joined: 28-October 11
Member No.: 94764



I think I've found the reason for the mentioned poor high frequency handling, at least in some cases. It seems to be simply a bug that content at the right beginning of a file is not encoded properly. If this "beginning" part is set somewhere else, everything's fine again. Can it be fixed somehow?
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 22nd August 2014 - 20:43